4 ms·
It would be interesting to see AlphaZero trained to play very aggressive romantic style. AlphaImmortal's goal should be beating Magnus Carlsen 51% of the time
by MAXPOOL 7y ago
It would be interesting to see AlphaZero trained to play very aggressive romantic style. AlphaImmortal's goal should be beating Magnus Carlsen 51% of the time with the most amazing games ever played.
I think you could do this by changing the value function:
- reward short game over long games,
- draw reward closer to loss reward,
- reward capture more.
In fact, I would like to play against AlphaImmortal specially trained to play against my Elo-rating level opponents with 51% win rate. That would be fun!
- Someone 7y ago"AlphaZero trained to play very aggressive romantic style. […] I think you could do this by changing the value function” The essential thing with AlphaZero is that it doesn’t have a value function that you can change. It learns it from the games it plays against itself. The best you could do is train it by playing against some algorithm that plays aggressive romantic style (playing against humans playing aggressive romantic style would work, too, but that would make for slow training) I doubt that would produce an AlphaZero that itself plays aggressive romantic style, though. Edit: one thing that _might_ work is to, during training, change the rules of chess to “both players lose if the game lasts over N moves” for some small N (40?). That should put incentives on aggressive play. Disadvantage would be that AlphaZero would learn very little about endgames.
- MAXPOOL 7y agoYou can make AlphaZero learn different value function by changing the reward from the game. Game with no reward from draw, more reward from shorter games with more captures creates different value functions.
- V-2 7y agoOr by training it against a flawed version of the algorithm that tends to overlook best defence like a human would. This was what made the romantic style possible in the first place: humans perform worse under pressure, so they tend to crack once they're put on the ropes. It's psychologically easier to attack than to defend accurately
- Someone 7y agoI don’t think that would work. A weird attacking style only works if your attacking moves that technically are losing ones lead to situations so complex that your opponent are unlikely to find the moves that win it for them. AlphaZero would not have trouble finding the (likely fairly dull) winning moves, and certainly wouldn’t see any reason to play such a move itself.
- pickdenis 7y agoThe chess computing term for this is "contempt," the parameter that determines the worst move the engine assumes its opponent would play. Classical chess engines can allow this parameter to be changed easily. I'm sure the MCTS algorithm for alphazero could be modified to allow this as well, if it doesn't already support it.
- V-2 7y agoAs a matter of fact, AlphaZero already plays more aggressively than typically expected of chess engines. Eg. in a well-known game it beated Stockfish (the strongest traditional chess engine), brilliantly sacrificing 3 pawns for a non-obvious, human-like compensation. This once again challenged the notion of what the perfect play boils down to in chess. Oversimplifying a bit, it had already been assumed that it's solely about grinding the opponent down by mercilessly accumulating miniscule advantages, while all the flashiness only worked because human players are error-prone. As it turns out, it may have been just another limitation of chess engines as we knew (and built) them.
- lonelappde 7y agoYou mean "reward capture less", to emphasize positional place and sacrifices.
- nabla9 7y agoAlphaZero against my level player using some wacky reward like sacrificing maximum amount of pawns over winning every time could wind me up like a God in bath salts.
- rictic 7y agoOnly somewhat related, but if you found these ideas amusing, you'd probably enjoy watching tom7's video on unusual chess engine strategies: https://youtu.be/DpXy041BIlA https://youtu.be/DpXy041BIlA
- freepor 7y agoAs a non-player who knows the rules these would be fun to watch. When I watch world championship matches I don’t really understand why what’s happening is happening. It would be cool to see games that are over optimized for the daring sacrifices that seem like blunders even to total novices but then end up being devious traps.